Reading the Recent Monetary History of the United States, 1959-2007

Jesús Fernández-Villaverde, Pablo Guerrón-Quintana, and Juan F. Rubio-Ramírez

In this paper the authors report the results of the estimation of a rich dynamic stochastic general equilibrium (DSGE) model of the U.S. economy with both stochastic volatility and parameter drifting in the Taylor rule. They use the results of this estimation to examine the recent monetary history of the United States and to interpret, through this lens, the sources of the rise and fall of the Great Inflation from the late 1960s to the early 1980s and of the Great Moderation of business cycle fluc - tuations between 1984 and 2007. Their main findings are that, while there is strong evidence of changes in monetary policy during Chairman Paul Volcker’s tenure at the Federal Reserve, those changes contributed little to the Great Moderation. Instead, changes in the volatility of structural shocks account for most of it. Also, although the authors find that monetary policy was different under Volcker, they do not find much evidence of a big difference in monetary policy among the tenures of Chairmen Arthur Burns, G. William Miller, and Alan Greenspan. The difference in aggre - gate outcomes across these periods is attributed to the time-varying volatility of shocks. The his - tory for inflation is more nuanced, as a more vigorous stand against it would have reduced infla tion in the 1970s, but not completely eliminated it. In addition, they find that volatile shocks (espe - cially those related to aggregate demand) were important contributors to the Great Inflation. (JEL E10, E30, C11)

Federal Reserve Board of St. Louis Review , July/August 2010, 92 (4), pp. 000-000.

1. INTRODUCTION process. First and foremost, documents are not a perfect photograph of reality. For example, par tici - ncovering the rationales behind mone tary pants at Federal Open Market Committee (FOMC) policy is hard. While the instruments of meetings do not necessarily say or vote what they Upolicy, such as the federal funds rate or really would like to say or vote, but what they reserve requirements, are directly observable, the think is appropriate at the moment given their process that led to their choice is not. Instead, objectives and their assessment of the strategic we have the documentary record of the minutes interactions among the members of the commit tee. of different meetings, the memoirs of partici pants (The literature on cheap talk and strategic voting in the process, and the internal memos circulat ed is precisely based on those insights.) Also, mem - within the Federal Reserve System. oirs are often incomplete or faulty and staff memos Although this paper trail is valuable, it is not are the product of negotiations and compromises and cannot be a complete record of the policy among several actors. Second, even the most com -

Jesús Fernández-Villaverde is an associate professor of economics at the University of Pennsylvania, a research associate for the National Bureau of Economic Research, a research affiliate for the Centre for Economic Policy Research, and a research associate chair for FEDEA (Fundación de Estudios de Economía Aplicada). Pablo Guerrón-Quintana is an economist at the Federal Reserve Bank of Philadelphia. Juan F. Rubio-Ramírez is an associate professor of economics at Duke University and a visiting scholar at the Federal Reserve Bank of Atlanta and FEDEA. The authors thank André Kurmann, Jim Nason, Frank Schorfheide, Tao Zha, and participants at several seminars for useful com ments, and Béla Személy for invaluable research assistance. The authors also thank the National Science Foundation for financial support. © 2010, The Federal Reserve Bank of St. Louis. The views expressed in this article are those of the author(s) and do not necessarily reflect the views of the Federal Reserve System, the Board of Governors, or the regional Federal Reserve Banks. Articles may be reprinted, reproduced, published, distributed, displayed, and transmitted in their entirety if copyright notice, author name(s), and full citation are included. Abstracts, synopses, and other derivative works may be made only with prior written permission of the Federal Reserve Bank of St. Louis.

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 1 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez plete documentary evidence cannot capture the a model that is structural in the sense of Hurwicz full richness of a policy decision process in a (1962)—that is, invariant to interventions such modern society. Even if it could, it would proba - as the ones that we consider. We need a model bly be impossible for any economist or historian with stochastic volatility because, otherwise, any to digest the whole archival record. 1 Third, even changes in the variance of aggregate variables if we could forget for a minute about the limita - would be interpreted as the consequence of vari - tions of the documents, we would face the fact ations in monetary policy. The evidence in Sims that actual decisions tell us only about what was and Zha (2006), Fernández-Villaverde and Rubio- done, but say little about what would have been Ramírez (2007), and Justiniano and Primiceri done in other circumstances. And while the (2008) points out that these changes in volatility absence of an explicit counterfactual may be a are first-order considerations when we explore minor problem for historians, it is a deep flaw the data. We need a model with parameter drift - for economists who are interested in evaluating ing in the monetary policy rule because we want policy rules and making recommendations regard - to introduce changes in policy that obey a fully ing the response to future events that may be very specified probability distribution, and not a once- different from past experiences. and-for-all change around 1979-80, as is often Therefore, in this paper we investigate the postulated in the literature (for example, in history of monetary policy in the United States Clarida, Galí, and Gertler, 2000, and Lubick and from 1959 to 2007 from a different perspective. Schorfheide, 2004). We build and estimate a rich dynamic stochastic In addition to using our estimation to inter - general equilibrium (DSGE) model of the U.S. pret the recent monetary policy history of the economy with both stochastic volatility and United States, we follow Sims and Zha’s (2006) parameter drifting in the Taylor rule that deter - call to connect estimated changes to historical mines monetary policy. Then, we use the results events. (We are also inspired by Cogley and of our estimation to examine, through the lens Sargent, 2002 and 2005.) In particular, we dis cuss of the model, the recent monetary policy history how our estimation results relate to both the obser - of the United States. Our attention is focused vations about the economy—for instance, how primarily on understanding two fundamental our model interprets the effects of oil shocks— observations: (i) the rise and fall of the Great and the written record. Inflation from the late 1960s to the early 1980s, the only significant peacetime inflation in U.S. Our main findings are that, although there is history, and (ii) the Great Moderation of business strong evidence of changes in monetary policy cycle fluctuations that the U.S. economy experi - during Chairman Paul Volcker’s tenure at the Fed, enced between 1984 and 2007, as documented those changes contributed little to the Great by Kim and Nelson (1998), McConnell and Pérez- Moderation. Instead, changes in the volatility Quirós (2000), and Stock and Watson (2003). of structural shocks account for most of it. Also, All the different elements in our exercise are although we find that monetary policy was dif fer - necessary. We need a DSGE model because we ent under Volcker, we do not find much evi dence are interested in counterfactuals. Thus, we require of a difference in monetary policy among the tenures of Chairmen Arthur Burns, G. William

1 For instance, Allan Meltzer (2010), in his monumental A History Miller, and Alan Greenspan. The reduction in the of the Federal Reserve , uses the summaries of the minutes of FOMC volatility of aggregate variables after 1984 is attrib - meetings compiled by nine research assistants (volume 2, book 1, uted to the time-varying volatility of shocks. The page X). This shows how even a several-decades-long commit ment to getting acquainted with the archives is not enough to process history for inflation is more subtle. According to all the relevant information. Instead, it is necessary to rely on sum - our estimated model, a more aggressive stance of maries, with all the potential biases and distortions that they might bring. This is, of course, not a criticism of Meltzer: He just pro ceeded, monetary policy would have reduced inflation as many other great historians do, by standing on the shoulders of in the 1970s, but not completely eliminated it. In others. Otherwise, modern archival research would be plainly impossible. addition, we find that volatile shocks (especially

2 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez those related to aggregate demand) were impor tant Unfortunately, for our purposes, the model contributors to the Great Inflation. has two weak points that we must acknowledge Most of the material in this paper is based before proceeding further: money and Calvo pric - on a much more extensive and detailed work by ing. Most DSGE models introduce a demand for Fernández-Villaverde, Guerrón-Quintana, and money through money in the utility function (MIA) Rubio-Ramírez (2010), in which we (i) present the or cash in advance (CIA). By doing so, w e endow DSGE model in all of its detail, (ii) characterize money with a special function without sound the decision rules of the agents, (iii) build the justification. This hides inconsistencies that are likelihood function, and (iv) estimate the model. difficult to reconcile with standard economic Here, we concentrate instead on understanding theory (Wallace, 2001). Moreover, the relation between structures wherein money is essential recent U.S. monetary history through the lens of and the reduced forms embodied by MIA or CIA our theory. is not clear. This means that we do not know whether that relation is invariant to changes in monetary policy or to the stochastic properties 2. A DSGE MODEL OF THE U.S. of the shocks that hit the economy, such as the ECONOMY WITH STOCHASTIC ones we study. This is nothing more than the VOLATILITY AND PARAMETER Lucas critique dressed in a different way. The second weakness of our DSGE model is DRIFTING the use of Calvo pricing. Probably the best way As we argued in the introduction, we need a to think about Calvo pricing is as a convenient structural equilibrium model of the economy to reduced form of a more-complicated pricing mech - evaluate the importance of each of the different anism that is easier to handle, thanks to its mem - mechanisms behind the evolution of inflation and oryless properties. However, if we are entertaining aggregate volatility in the United States over the the idea that monetary policy or the volatility of past several decades. However, while the previ ous shocks has changed over time, it is exceedingly statement is transparent, it is much less clear how difficult to believe that the parameters that con trol to decide which particular elements of the model Calvo pricing have been invariant over the same to include. On the one hand, we want a model period (see the empirical evidence that supports that is sufficiently detailed to account for the this argument in Fernández-Villaverde and Rubio- dynamics of the data reasonably well. But this goal Ramírez, 2008). conflicts with the objective of having a parsimo - However, getting around these two limita tions nious and soundly microfounded description of seems, at the moment, infeasible. Microfounded models of money are either too difficult to work the aggregate economy. with (Kiyotaki and Wright, 1989) or rest in assump - Given our investigation, a default choice for tions nearly as implausible as MIA (Lagos and a model is a standard DSGE economy with nom i- Wright, 2005) or that the data find too stringent nal rigidities, such as the ones in Christiano, (Aruoba and Schorfheide, forthcoming). State- Eichenbaum, and Evans (2005) or Smets and dependent models of pricing are too cumber some Wouters (2003). This class of models is currently computationally for estimation (Dotsey, King, and being used to inform policy in many central banks Wolman, 1999). and is a framework that has proven to be suc cess - So, with a certain reluctance, we use a main - ful at capturing the dynamics of the data. How - stream DSGE model with households; firms (a ever, we do not limit ourselves to a standard DSGE labor packer, a final-good producer, and a con tin - model. Instead, we extend it in what we think uum of intermediate-good producers); a mone tary are important and promising directions by incor - authority, the Federal Reserve, which imple ments porating stochastic volatility into the structural monetary policy through open market opera tions shocks and parameter drifting in the Taylor rule following a Taylor rule; and nominal rigidities in that governs monetary policy. the form of Calvo pricing with partial indexa tion.

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 3 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

where is the conditional expectation operator, 2.1 Households ޅ0 β is the discount factor for one quarter (the time We begin our discussion of the model with period for our model), h controls habit persist ence, households. We work with a continuum of them, and ϑ is the inverse of the Frisch labor supply indexed by j. Households are different because elasticity. In addition, we introduce two shifters each supplies a specific type of labor in the mar - to preferences, common to all households: The ket: Some households are carpenters and some first is a shifter to intertemporal preference, d , households are economists. If, in addition, each t that makes utility today more or less desirable. household has some market power over its own This is a simple device to capture shocks to aggre - wage and stands ready to supply any amount of labor at posted prices, it is relatively easy to intro - gate demand. A prototypical example could be duce nominal rigidities in wages. Some house - increases in aggregate demand caused by fiscal holds are able to change their wages and some policy, an aspect of reality ignored in our model. Another possibility is to think about d as the con - are not, and the relative demand for each type of t labor adjusts to compensate for these differences sequence of demographic shocks that propagate in input prices. over time. The second is a shifter to labor sup ply, At the same time, we do not want a compli - ϕt. As emphasized by Hall (1997), this shock is cated model with heterogeneous agents that is crucial for capturing the fluctuation of hours in daunting to compute. We resort to two tricks to the data. get around that problem. First, we have a utility A simple way to parameterize the evolution function that is separable among consumption, of the two shifters is to assume AR(1) processes: cjt, real money balances, mjt/pt, and hours worked, logdt =+ρσεdlogdt−1 dt dt , ljt. Second, we have complete markets in Arrow securities. Complete markets allow us to equate where εdt ~ N͑0,1 ͒, and the marginal utilities of consumption across all households in all states of nature. And, since by logϕρt =+ϕϕϕlog ϕt−1 σεt t , separability this marginal utility depends only on where ε ~ 0,1 . The most interesting feature consumption, all households will consume the ϕt N͑ ͒ of these processes is that the standard deviations same amount of the final good. The result makes (SDs), σ and σ , of the innovations, ε and ε , aggregation trivial. Of course, it also has the dt ϕt dt ϕt unpleasant feature that those households that do evolve over time. This is the first place where we not update their wages will work different num - introduce time-varying volatility in the model: bers of hours than those that do. If, for example, Sometimes the preference shifters are highly we have an increase in the average wage, those volatile; sometimes they are less so. This chang - households stuck with the old, lower wages will ing volatility may reflect, for instance, the differ - work longer hours and have lower total utility. ent regimes of fiscal policy or the consequences This is the price we need to pay for tractability. of demographic forces (Jaimovich and Siu, 2009). Given our previous choice of a separable We can specify many different processes for utility function and our desire to have a bal anced σdt and σϕt. A simple procedure is to assume that growth path for the economy (which requires a σdt and σϕt follow a Markov chain and take a finite marginal rate of substitution between labor and number of values. While this specification seems consumption that is linear in consumption), we straightforward, it is actually quite involved. The postulate a utility function of the form distribution that it implies for σdt and σϕt is dis - crete and, therefore, perturbation methods (such   log(cjt − hcjt−1 ) as the ones that we use later) are ill designed to ∞   (1) ޅ ∑βtd  1+ϑ , deal with it. Such conditions would force us to 0 t  mjt  ljt t=0 +υlog − ϕ ψ  rely on global solution methods that are too slow  p  t 1+ ϑ   t   for estimation.

4 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Instead, we can postulate simple AR(1) proc - where εµt ~ N͑0,1 ͒ with drift Λµ and innovation esses in logs (to ensure the positivity of the SDs): εµt, whose SD σµt evolves according to our favorite autoregressive process: logσρσρση= ()1− log ++log u , dt σσd d d dt−1 d dt

logσρσρσηµσµσµµµt = ()1− log ++log t−1 u t , where udt ~ N͑0,1 ͒, and µµ

where uµt ~ N͑0,1 ͒. logσρσρσηϕσϕσϕϕϕt = ()1− log ++log t−1 u t , ϕϕ We introduce this shock convinced by the where uϕt ~ N͑0,1 ͒. This specification is both evidence in Greenwood, Herkowitz, and Krusell parsimonious (with only four new parameters, (1997) that this is a key mechanism to under stand - ing aggregate fluctuations in the United States ρσ , ρσ , ηd, and ηϕ ) and rather flexible. Because d ϕ of these advantages, we impose the same specifi - over the past 50 years. cation for the other three time-varying SDs in the Thus, the jth household’s budget constraint is model that appear below (the ones affecting an mjt bjt+1 cjt ++x jt + +∫qjt++1,tajt 1dω jt +1,t investment-specific technological shock, a neu tral p p technology shock, and a monetary policy shock). t t Hereafter, agents perfectly observe the structural = w l +−r u µ −1Φu  k jt jt ()t jt t  jt  jt−1 shocks and the level and innovation to the SDs m b and have rational expectations about their sto - jt−1 jt ++Rt−1 + ajt ++Tt Ft , chastic properties. pt pt Households keep a rich portfolio: They own where wjt is the real wage, rt is the real rental (physical) capital, kjt; nominal government bonds, price of capital, ujt > 0 is the rate of use of capi tal, bjt, that pay a gross return Rt–1; Arrow securities, –1 µt Φ[ujt] is the cost of using capital at rate ujt in ajt+1 , which pay one unit of consumption in event terms of the final good, µt is an investment-spe cific ω traded at time t at unitary price q ; and jt+1, t jt+1, t technology level, Tt is a lump-sum transfer, and cash. Ft is the profits of the firms in the economy. We The evolution of capital deserves some postulate a simple quadratic form for Φ[.], description. Given a depreciation rate δ, the Φ amount of capital owned by household j at the 2 2 ΦΦ[]u =−1 ()u 1 +−()u 1 , end of period t is 2 and normalize u, the utilization rate in the bal -    x jt  anced growth path of the economy, to 1. This k =−()1 δµk +−1 V   x . jt jt−1 t  x  jt imposes the restriction that the parameter Φ must   jt−1  1 satisfy Φ1 = Φ′ [1] = r˜, where r˜ is the balanced Investment, xjt, is multiplied by a term that growth path rental price of capital (rescaled by depends on a quadratic adjustment cost func tion, technological progress, as we explain later).

2 Of all the choice variables of the households,     xt κ xt the only one that requires special attention is V   =− Λx  ,  xt−−1  2  xt 1  hours. As we explained previously, each house - hold j supplies its own specific type of labor. written in deviations with respect to the bal anced This labor is aggregated by a labor packer into growth rate of investment, Λx, with adjustment d homogeneous labor, lt , according to a constant parameter κ and an investment-specific technol - elasticity of substitution technology, ogy level µ . This technology level evolves as a t η random walk in logs:  η−1  η−1 d 1 η lt =  ∫ l jt dj .  0  logµµσεt =+Λµµµlog t−1 +t t ,  

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 5 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

The labor packer is perfectly competitive and logAt =+ΛA logAt−1 +σεAt At , takes all the individual wages, wjt, and the wage d where ε ~ 0,1 with drift Λ and innovation ε . wt for lt as given. At N͑ ͒ A At The household decides, given the demand We keep the same specification for the SD of this function for its type of labor generated by the innovation as we did for all previous volatilities: labor packer, logσρσρση= ()1− log ++log u , At σ A A σ A At −1 A At w −η  jt  d l jt = lt ∀j, where u ~ N 0,1 .  w  At ͑ ͒  t  The quantity sold of the good is determined which wage maximizes its utility and stands ready by the demand function (equation (3)). Given to supply any amount of labor at that wage. How - equation (3), the intermediate-good producers ever, when it chooses the wage, the household is set prices to maximize profits. As was the case subject to a nominal rigidity: a Calvo pricing for households, intermediate-good producers are mechanism with partial indexation. At the start subject to a nominal rigidity in the form of Calvo pricing. In each quarter, a proportion of them, of every quarter, a fraction 1 – θw of households are randomly selected and allowed to reoptimize 1 – θp, can reoptimize their prices. The remain ing their wages. All other households can only index fraction θp indexes their prices by a fraction their wages to past inflation with an indexation χ ʦ [0,1] of past inflation. parameter χw ʦ [0,1]. 2.3 The Policy Rule of the Federal 2.2 Firms Reserve In addition to the labor packer, we have two In our model, the Federal Reserve imple ments other types of firms in this economy. The first, the monetary policy through open market opera tions final-good producer, is a perfectly competitive (that generate lump-sum transfers, Tt, to main tain firm that aggregates a continuum of intermediate a balanced budget). In doing so, the Fed follows goods with the technology: a modified Taylor rule that targets the ratio of nominal gross return, R , of government bonds ε t 1 ε −1 ε −1 d ε over the balanced growth path gross return, R: (2) yt = ()∫ y it di . 0 1 γ −γ R  y ,t  γγ yt  This firm takes as given all intermediate-good R R R Π Π,t t  t−1   t  yt −1  prices, p , and the final-good price, p , and gen - =       ξt . ti t R  R   Π   exp()Λ   erates a demand function for each intermediate   y   good: This rule depends on (i) the past Rt–1, which −ε  p  smooths changes over time; (ii) the “inflation gap,” (3) y = it y d ∀i. it   t Π /Π, where Π is the balanced growth path of  pt  t inflation 2; (iii) the “growth gap,” which is the ratio Second, we have the intermediate-good pro - ducers, each of which has access to a Cobb- 2 Here we are being careful with our words: Π is inflation in the Douglas production function, balanced growth path, not the target of inflation in the stochastic steady state. As we will see later, we solve the model using a sec ond- α d 1−α order approximation. The second-order terms move the mean of y it = At kit−1 ()lit , the ergodic distribution of inflation, which corresponds in our view to the usual view of the inflation target, away from the balanced where k is the capital, ld is the packed labor growth path level. We could have expressed the policy rule in terms it–1 it of this mean of the ergodic distribution, but it would amount to rented by the firm, and At (our fourth structural solving a complicated fixed-point problem (for every inflation level, shock) is the neutral productivity level, which we would need to solve the model and check that indeed this is the mean of the ergodic distribution), which is too complicated a evolves as a random walk in logs: task for the potential benefits we can derive from it.

6 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez between the growth rate of the economy, yt /yt–1, been particularly true in the late 1960s, when a and Λy, the balanced path gross growth rate of yt, majority of staff economists embraced Keynesian dictated by the drifts of neutral and investment- policies and the MIT-Penn-Federal Reserve System specific technological change; and (iv) a mone tary (MPS) model was built. 4 The second phenome non σm,tεmt policy shock, ξt = exp , with an innova tion is the observation that, even if we keep constant εmt ~ N͑0,1 ͒ and SD of the innovation, σm,t, that the members of the FOMC, their reading of the evolves as priorities and capabilities of monetary policy may evolve (or be more or less influenced by the gen - logσρσρση= ()1− log ++log u . mt σm m σ m mt−1 m m,t eral political climate of the nation). We argue below that this is a good description of Martin, Note that, since we are dealing with a general who changed his beliefs about how strongly the equilibrium model, once the Fed has chosen a Fed could fight inflation in the late 1960s, or value of Π, R is not a free target, as it is deter mined Greenspan’s growing conviction in the mid- 1990s by technology, preferences, and Π. that the long-run growth rate of the U.S. econ omy We introduce monetary policy changes had risen. through a parameter drift over the responses of While this second channel seems well R to inflation, γ , and growth gaps, γ : t Π,t y,t described by a continuous drift in the parame ters (beliefs plausibly evolving slowly), changes in logγργργηεΠΠΠt =−()1 γγlog ++log t−1 ππt , ΠΠ the voting members, in particular the chairman, where επt ~ N͑0,1 ͒ and might potentially be better understood as dis crete

jumps in γ Π,t and γy,t. In fact, our smoothed path logγργργηεyt =−()1 γγlog y ++log yt −1 y yt , y y of γ Π,t, which we estimate from the data, gives some support to this view. But in addition to our where ε ~ 0,1 . yt N͑ ͒ pragmatic consideration that computing models In preliminary estimations, we discovered with discrete jumps is hard, we argue in Section 6 that, while other parameters such as γ could also R that, historically, changes have occurred more be changing, the likelihood function of the model slowly and even new chairmen have required did not react much to that possibility and thus, some time before taking a decisive lead on the we eliminated those channels. FOMC (Goodfriend and King, 2005). Our parameter-drifting specification tries to In Section 7, we discuss other objections to capture mainly two different phenomena. First, our form of parameter drifting—in particular the changes in the composition of the voting mem bers objection to the assumption that agents observe of the FOMC (through changes in governors and the changes in parameters without problem, its in the rotating votes of presidents of regional exogeneity, or its avoidance of open economy Reserve Banks) may affect how strongly the FOMC considerations. responds to inflation and output growth because of variations in the political-economic equilib rium 2.4 Aggregation and Equilibrium in the committee. 3 Similarly, changes in staff may have effects as long as their views have an impact The model is closed by finding an expres sion on the voting members through briefings and for aggregate demand, other, less-structured interactions. This may have d −1 yt =++ct xt µt Φut  kt −1,

3 According to Walter Heller, President Kennedy clearly stated, and another for aggregate supply, “About the only power I have over the Federal Reserve is the power of appointment, and I want to use it” (cited by Bremner, 2004, p. 160). The slowly changing composition of the Board of Governors 4 The MPS model is the high-water mark of traditional Keynesian may lead to situations, such as the one in February 1986 (dis - macroeconometric models in the Cowles tradition. The MPS model cussed later in the text), when Volcker was outvoted by Ronald was used operationally by staff economists at the Fed from the Reagan’s appointees on the Board. early 1970s to the mid-1990s (see Brayton et al., 1997).

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 7 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

and our investigation would be essentially s 1 α d 1−α yt = At ()ut kt−1 ()lt , worthless. v p t Nearly by default, then, using perturbation where to obtain a higher-order approximation to the equilibrium dynamics of our model is the only d 1 1 lt = ∫ ljtdj option. A second-order approximation includes v w 0 t terms that depend on the level of volatility. Thus, is demanded labor, these terms capture the responses of agents (house -

−η holds and firms) to changes in volatility. At the 1  w  v w = jt dj same time, a second-order approximation can be t ∫0    wt  found sufficiently fast, which is of the utmost importance since we want to estimate the model is the aggregate loss of labor input induced by and that forces us to solve it repeatedly for many wage dispersion, and different parameter values. Thus, a second-order −ε approximation is an interesting compromise p 1  pit  vt = ∫ di between accuracy and speed. 0 p  t The idea of perturbation is simple. Instead of is the aggregate loss of efficiency induced by price the exact decision rule of the agents in the model, dispersion of the intermediate goods. By market we use a second-order Taylor expansion to the d s clearing, yt = yt = yt . rule around the steady state. That Taylor expan - The definition of “equilibrium” for this model sion depends on the state variables and on the is rather standard: It is just the path of aggregate innovations. However, we do not know the coef - quantities and prices that maximize the prob lems ficients multiplying each term of the expansion. of households and firms, the government fol lows Fortunately, we can find them by an application its Taylor rule, and markets clear. But while the of the implicit function theorem as follows (see definition of equilibrium is straightforward, its also Judd, 1998, and Schmitt-Grohé and Uribe, computation is not. 2006). First, we write all the equations describing the equilibrium of the model (optimality conditions 3. SOLUTION AND LIKELIHOOD for the agents, budget and resource constraints, the Taylor rule, and the laws of motion for the EVALUATION different stochastic processes). Second, we rescale The solution of our model is challenging. We all the variables to remove the balanced growth have 19 state variables, 5 innovations to the struc - path induced by the presence of the drifts in the tural shocks ( εdt , εϕt, εAt , εµt, εmt), 2 innovations to evolution of neutral and investment-specific tech - the parameter drifts ( επt, εyt), and 5 innovations to nology. Third, we find the steady state implied by the volatility shocks ( udt , uϕt, uµt, uAt , umt), for a the rescaled variables. Fourth, we linearize the total of 31 variables that we must consider. equilibrium conditions around the steady state A vector of 19 states makes it impossible to found in the previous step. Then, we solve for the use value-function iteration or projection meth ods unknown coefficients in this linearization, which (finite elements or Chebyshev polynomials). The happen to be, by the implicit function theorem, curse of dimensionality is too acute even for the the coefficients of the first-order terms of the deci - most powerful of existing computers. Standard sion rules in the rescaled variables that we are linearization techniques do not work either: looking for (which can be easily rearranged to Stochastic volatility is inherently a nonlinear deliver the decision rules in the original vari ables). process. If we solved the model by linearization, The next step is to take a second-order approxi - all terms associated with stochastic volatility mation of the equilibrium conditions, plugging would disappear because of certainty equiva lence in the terms found before, and solve for the coef -

8 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez ficients of the second-order terms of the decision Now that we have an evaluation of the likeli - rules. hood of the model given observables, we only While we could keep iterating in this proce - need to search over different parameter values dure for as long as we want, Aruoba, Fernández- according to our favorite estimation algorithm. Villaverde, and Rubio-Ramírez (2006) show that, This can be done in two ways. One is with a reg - for the basic stochastic neoclassical growth model ular maximum likelihood algorithm: We look for (the backbone of our model) calibrated to the U.S. a global maximum of the likelihood. This proce - data, a second-order approximation delivers excel - dure is complicated by the fact that the evalua tion lent accuracy at great computational speed. In our of the likelihood function that we obtain from the actual computation, we undertake the symbolic particle filter is nondifferentiable with respect to derivatives of the equilibrium conditions using the parameters because of the inherent discrete - Mathematica 6.0. The code generates all of the ness of the resampling step. An easier alterna tive, relevant expressions and exports them automat - and one that allows the introduction of presam ple ically into Fortran files. Then, Fortran sends information, is to follow a Bayesian approach. In par ticular parameter values in each step of the this route, we specify a prior over the parame ters, estimation, evaluates those expressions, and multiply the likelihood by it, and sample from the determines the terms of the Taylor expansions resulting posterior by means of a random-walk that we need. Metropolis-Hastings algorithm. In this paper, we Once we have the approximated solution to choose this second route. In our estimation, how- the model, given some parameter values, we use ever, we do not take full advantage of presample it to build a state-space representation of the information since we impose flat priors to facili - dynamics of states and observables. This repre - tate the communication of the results to other sentation is, as we argued before, nonlinear and researchers: The shape of our posterior distribu - hence standard techniques such as the Kalman tions will be proportional to the likelihood. We filter cannot be applied to evaluate the associ ated must note, however, that relying on flat priors likelihood function. Instead, we resort to a simu - forces us to calibrate some parameters to values lation method known as the particle filter, as typically used in the literature (see Fernández- applied to DSGE models by Fernández- Villaverde Villaverde, Guerrón-Quintana, and Rubio-Ramírez, and Rubio-Ramírez (2007). The particle filter gen - 2010 [FGR hereafter], for the values and justifi ca - erates a simulation of different states of the model tion of the calibrated values). and evaluates the probability of the innovations While our description of the solution and that make these simulated states explain the estimation method has been necessarily brief, observables. These probabilities are also called the reader is invited to check FGR for additional weights. A simple application of a law of large details. In particular, FGR characterize the struc - numbers tells us that the mean of the weights is ture of the higher-order approximations, show ing an evaluation of the likelihood. The secret of the that many of the relevant terms are zero and success of the procedure is that, instead of per - exploiting this result to quickly solve for the inno - forming the simulation over the whole sample, vations that explain the observables given some we perform it only period by period, resampling states. This result, proved for a general class of from the set of simulated state variables accord - DSGE models with stochastic volatility, is bound to ing to the weights we just found. This sequential have wide application in all cases where sto chas - structure, which makes the particle filter a case of tic volatility is an important aspect of the prob lem. a more general class of algorithms called sequen - tial Monte Carlo methods, ensures that the simu - lation of the state variables remains centered on the true but unknown value of the state vari ables. 4. ESTIMATION This dramatically limits the numerical variance To estimate our model, we use five time series of the procedure. for the U.S. economy: (i) the relative price of

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 9 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Table 1 Posterior Distributions: Parameters of the Stochastic Processes for Volatility Shocks

Parameter σ σ σ σ σ log d log ϕ log µ log A log m –1.9834 –2.4983 –6.0283 –3.9013 –6.000 (0.0726) (0.0917) (0.1278) (0.0745) (0.1471)

ρσ ρσ ρσ ρσ ρσ d ϕ µ a m 0.9506 0.1275 0.7508 0.2411 0.8550 (0.0298) (0.0032) (0.035) (0.005) (0.0231) η η η η η d ϕ µ a m 0.3246 2.8549 0.4716 0.7955 1.1034 (0.0083) (0.0669) (0.006) (0.013) (0.0185)

NOTE: Numbers in parentheses indicate standard deviations. investment goods with respect to the price of similar to inflation: It rises in the 1970s (although consumption goods, (ii) the federal funds rate, less than inflation during the earlier years of the (iii) real per capita output growth, (iv) the con - decade and more during the last years) and stays sumer price index, and (v) real wages per capita. much lower in the 1990s, reaching historical Our sample covers 1959:Q1 to 2007:Q1. minima by the end of the sample. Figure 1 plots three of the five series: infla tion, The point estimates we get from our poste rior (per capita) output growth, and the federal funds distribution agree with other estimates in the lit - rate—the three series most commonly discussed erature. For example, we document a fair amount when commentators talk about monetary policy. of nominal rigidities in the economy. In any case, By refreshing our memory about their evolution we refer the reader to FGR to avoid a lengthy dis - in the sample, we can frame the rest of our dis - cussion. Here, we report only the modes and SDs cussion. For ease of reading, each vertical bar of the posterior distributions associated with the corresponds to the tenure of one Fed chairman: parameters governing stochastic volatility (Table 1) Martin, Burns-Miller (we merge these two because and policy (Table 2). In our view, those parame - of Miller’s short tenure), Volcker, Greenspan, and ters are the most relevant for our reading of the Bernanke. recent history of monetary policy in the United The top panel shows the history of the Great States. Inflation: From the late 1960s to the mid-1980s, The main lesson from Table 1 is that the scale the U.S. experienced its only significant infla tion parameters, ηi, are clearly positive and bounded in peace time, with peaks of around 12 to 14 per - away from zero, confirming the presence of time- cent during the 1973 and 1979 oil shocks. The variant volatility in the data. Shocks to the volatil - middle panel shows the Great Moderation: A sim - ity of the intertemporal preference shifter, σd, are ple inspection of the series after 1984 reveals a the most persistent (also, the SDs are tight enough much smaller amplitude of fluctuations (espe - to suggest that we are not suffering from serious cially between 1993 and 2000) than before that identification problems). The innovations to the date. The Great Inflation and the Great Moderation volatility of the intratemporal labor shock, ηϕ , are the two main empirical facts to keep in mind are large in magnitude, which suggests that labor for the rest of the paper. The bottom panel shows supply shocks may have played an important role the federal funds rate, which follows a pattern during the Great Inflation by moving the mar ginal

10 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Figure 1 Time Series for Inflation, Output Growth, and the Federal Funds Rate

Annualized Rate Inflation 14 Martin 12 Burns-Miller 10 Volcker 8 Greenspan Bernanke 6

4

2

0

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

Annualized Rate Output 10

5

0

−5

−10

−15 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

Annualized Rate Interest Rate 18 16 14 12 10 8 6 4 2

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 11 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Table 2 Posterior Distribution: Policy Parameters

Parameter γ γ Π γ η R log y log Π π 0.7855 –1.4034 1.0005 0.0441 0.1479 (0.0162) (0.0498) (0.0043) (0.0005) (0.002)

NOTE: Numbers in parentheses indicate standard deviations. cost of intermediate-good producers. Finally, the inflation plus or minus a 2-SD interval given our estimates for the volatility process governing point estimates of the structural parameters. The investment-specific productivity suggest that message of Figure 2 is straightforward. According such shocks are important in accounting for to our model, at the arrival of the Kennedy admin - business cycles fluctuations in the United States istration, the response of monetary policy to (Fisher, 2006). inflation was around its estimated mean, slightly The results from Table 2 indicate that the over 1. 6 It grew more or less steadily during the central bank smooths interest rates ( γR > 0). The 1960s, until reaching a peak at the end of 1967 parameter γ Π is the average magnitude of the and beginning of 1968. Subsequently, γ Πt fell so response to inflation in the Taylor rule. Its esti - quickly that it was below 1 by 1971. For nearly mated value (1.045 in levels) is just enough to all of the 1970s, γ Πt stayed below 1 and picked up guarantee determinacy in the model (Woodford, only with the arrival of Volcker. Interestingly, the 2003). 5 The size of the innovations to the drift - two oil shocks did not have an impact on the esti - mated γ . The parameter stayed high through out ing inflation parameter, ηπ , reaffirms our view of Πt a time-dependent response to inflation in mone - the Volcker years and fell after a few quarters into Greenspan’s tenure, when it returned to levels tary policy. The estimates for γy,t (the response to output deviations in the Taylor rule) are not even lower than during the Burns and Miller years. reported because preliminary attempts at estima - The likelihood function favors an evolving mon e- tary policy even after introducing stochastic tion convinced us that ηy was nil. Hence, in our volatility in the model. In FGR, we assess this next exercises, we set ργ and ηy to zero. y statement more carefully with several measures of model fit, including the construction of Bayes 5. TWO FIGURES factors and the computation of Bayesian infor - mation criteria between different specifications In this section, we present two figures that of the model. show us much about the evolution and effects of The reader could argue, with some justifica - monetary policy: (i) the estimated smoothed path tion, that we have estimated a large DSGE model of γ over our sample and (ii) the evolution dur - Πt and that it is not clear what is driving the results ing the same years of a measure of the real inter est and what variation in the data is identifying the rate. In the next section, we map these figures into movements in monetary policy. While a fully the historical record. worked-out identification analysis is beyond the Figure 2, perhaps the most important figure scope of this paper, as a simple reality check, we in this paper, plots the smoothed estimate of the plot in Figure 3 a measure of the (short-term) evolution of the response of monetary policy to

6 This number nearly coincides with the estimate of Romer and 5 In this model, local determinacy depends only on the mean of γΠ. Romer (2002a) of the coefficient using data from the 1950s.

12 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Figure 2 Smoothed Path for the Taylor Rule Parameter on Inflation ±2 SDs

Level of Parameter Martin 5 Burns-Miller Volcker 4 Greenspan Bernanke

3

2

1

0

–1 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

Figure 3 Real Interest Rate (Federal Funds Rate Minus Inflation)

Annualized Rate

12 Martin Burns-Miller 10 Volcker Greenspan 8 Bernanke

6

4

2

0

–2

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 13 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez real interest rate defined as the federal funds rate Greenspan—except for Miller, whom we group minus current inflation. 7 with Burns because of his short tenure. This figure shows that Martin kept the real One fundamental lesson from this exercise is interest rate at positive values around 2 percent that Figure 2 can successfully guide our inter pre - during the 1960s (with a peak by the end, which tation of policy from 1959 to 2007. We document corresponds with the peak of our estimated γ Πt). how both Martin and Volcker believed that infla - However, during the 1970s, the real interest rate tion was dangerous and that the Fed had both the was often negative and only rarely above 2 per cent, responsibility and the power to fight it, although a rather conservative lower bound on the bal anced growing doubts about that power overcame Martin growth real interest rate given our point esti mates. during his last term as chairman. Burns, on the The likelihood function can interpret those obser - other hand, thought the costs of inflation were vations only as a very low γ Πt (remember that the lower than the cost of a recession triggered by Taylor principle calls for increases in the real disinflation. In any case, he was rather skeptical interest rate when inflation rises; that is, nomi nal about the Fed’s ability to successfully disinflate. interest rates must grow more than inflation). Real Greenspan, despite his constant warnings about interest rates skyrocketed with the arrival of inflation, had in practice a much more nuanced Volcker, reaching a historic record of 13 percent attitude. According to our estimated model, good by 1981:Q2. After that date, they were never even positive shocks to the economy gave him the close to zero, and only in two quarters were they privilege of skipping a daunting test of his resolve. below 3 percent. Again, the likelihood function Because by using a DSGE model we have a can interpret that observation only as a high γ Πt. complete set of structural and volatility shocks, The Greenspan era is more complicated because in FGR, we complete this analysis with the con - real interest rates were not particularly low in the struction of counterfactual exercises. In those exer - 1990s. However, output growth was very posi tive, cises, we build artificial histories of economies which pushed the interest rates up in the Taylor in which some source of variation has been elimi - rule. Since the federal funds rate was not as high nated or modified in an illustrative manner. For as the policy rule would have predicted with a example, we can evaluate how the economy high γ Πt, the smoothed estimate of the parameter would have behaved in the absence of changes is lowered. During the 2000s, real interest rates in the volatility of the structural shocks or if the close to zero were enough, by themselves, to keep average monetary policy of one period had been

γ Πt low. applied in another. By interpreting those coun ter - factual histories, we attribute most of the defeat of the Great Inflation to monetary policy under 6. READING MONETARY HISTO RY Volcker and most of the Great Moderation after THROUGH THE LENS OF OUR 1984 to good shocks. We incorporate informa tion from those counterfactuals as we move along. MODEL Our exercise in this section is closely related Now that we have our model and our esti - to the work of Christina and David Romer (1989; mates of the structural parameters, we smooth 2002a,b; 2004), except that we attack the prob lem the structural and volatility shocks implied by from exactly the opposite perspective. While they the data and use them to read the recent mone tary let their narrative approach guide their empiri cal history of the United States. Somewhat conven - specification and like to keep a flexible relation tionally, we organize our discussion around the with equilibrium models, we start from a tightly different chairmen of the Fed from Martin to parameterized DSGE model of the U.S. economy and use the results of our estimation to read the 7 Since inflation is nearly a random walk (Stock and Watson, narrative told by the documents. We see both 2007), its current value is an excellent proxy for its expected value. strategies as complementary since each can teach In any case, our argument is fully robust to slightly different defi - nitions of the real interest rate. us much of interest. Quite remarkably, given the

14 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez different research designs, many of our conclu - use of credit in line with resources available for sions are similar to the views expressed by Romer production of goods and services.” 9 Martin was and Romer. also opposed to the idea (popular at the time) that the U.S. economy had a built-in bias toward 6.1 The Martin Era: Resistance and inflation, a bias the Fed had to accommodate Surrender through monetary policy. Sumner Slichter, an influential professor of economics at Harvard, William McChesney Martin, the chairman of was perhaps the most vocal proponent of the the Fed between April 2, 1951, and January 31, built-in bias hypothesis. In Martin’s own words, 1970, knew how to say no. On December 3, 1965, “I refuse to raise the flag of defeatism in the bat tle he dared to raise the discount rate for the first time of inflation” and “there is no validity whatever in more than five years, despite warnings from in the idea that any inflation, once accepted, can the Treasury secretary, Henry Fowler, and the be confined to moderate proportions.” 10 As we chairman of the Council of Economic Advisors, will see in the next subsection, this opposition Gardner Ackley, that President Lyndon Johnson stands in stark contrast to Burns’s pessimistic disapproved of such a move. Johnson, a man not view of inflation, which had many points of con - used to seeing his orders ignored, was angered tact with Slichter’s. by Martin’s unwelcome display of independence Our estimates of γ Π, t, above 1 and growing and summoned him to a meeting at his Texas during the period, clearly tell us that Martin was ranch. There, for over an hour, he tried to corner doing precisely that: working to keep inflation low. the chairman of the Fed with the infamous bul - Our result also agrees with Romer and Romer’s lying tactics that had made him a master of the (2002a) narrative and statistical evidence regard - Senate in years past. Martin, however, held his ing the behavior of the Fed during the late 1950s. ground and carried the day: The raise would stand. We must not forget, however, that our estimates Robert Bremner starts his biography of Martin in FGR suggest as well that the good perform ance 8 with this story. The choice is most appropriate. of the economy from 1961 to 1965 was also the The history of this confrontation illustrates bet ter consequence of good positive shocks. than any other event our econometric results. The stand against inflation started to be tested The early 1960s were the high years of around 1966. Intellectually, more and more voices Martin’s tenure. The era of the “New Economics” had been raised since the late 1950s defending combined robust economic growth, in excess of the notion that an excessive concern with infla - 5 percent, and low inflation, below 3 percent. tion was keeping the economy from working at According to our estimated model, this moder ate full capacity. Bremner (2004, p. 138) cites Walter inflation was, in part, a reflection of Martin’s views Heller and Paul Samuelson’s statements before about economic policy. Bremner (2004, p. 122) the Joint Economic Committee in February 1959 summarizes Martin’s guiding principles this way: as examples of an attitude that would soon gain Stable prices were crucial for the correct work ing strength. The following year, Samuelson and of a market economy and the Fed’s main task was Robert Solow’s (1960) classic paper about the to maintain that stability. In Martin’s own words, Phillips curve was taken by many as providing “the Fed has a responsibility to use the powers it possesses over economic events to dampen 9 Martin’s testimony to the Joint Economic Committee, February 5, excesses in economic activity [by] keeping the 1957 (cited by Bremner 2004, p. 123).

10 The first quotation is from the New York Times , March 16, 1957, 8 Bremner (2004, pp. 1-2). This was not the only clash of Martin with where Martin was expressing dismay for having reached a 2 per cent a president of the United States. In late 1952, Martin bumped into rate of inflation. The second quotation is from the Wall Street Harry Truman leaving the Waldorf Astoria Hotel in New York City. Journal , August 19, 1957. Martin also thought that Keynes him self To Martin’s “Good afternoon,” Truman wryly replied, “Traitor!” had changed his views on inflation after the war (they had talked Truman was deeply displeased by how the Fed had implemented privately on several occasions) and that, consequently, Keynesian the accord of March 3, 1951, between the Fed and the Treasury that economists were overemphasizing the benefits of inflation. See ended the interest rate peg in place since 1942 (Bremner, 2004, p. 91). Bremner (2004, pp. 128 and 229).

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 15 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez an apparently sound empirical justification for egy worked. Heller’s first choice, George W. a much more sanguine position with respect to Mitchell, would become a leader of those prefer - inflation: “In order to achieve the nonperfec - ring a more expansionary monetary policy on tion ist’s goal of high enough output to give us the FOMC. no more than 3 percent unemployment, the price By 1964, Martin was considerably worried index might have to rise by as much as 4 to 5 per - about inflation. He told Johnson: “I think we’re cent per year. That much price rise would seem heading toward an inflationary mess that we won’t to be the necessary cost of high employment and be able to pull ourselves out of.” 13 In 1965, he production in the years immediately ahead” ran into serious problems with the president, as 11 (Samuelson and Solow, 1960, p. 192). Heller’s discussed at the beginning of this section. The and Tobin’s arrival on the Council of Economic problems appeared again in 1966 with the appoint - Advisors transformed the critics into the insid ers. ment of Brimmer as a governor against Martin’s The pressures on monetary policy were con - recommendation. During all this time, Martin tained during Kennedy’s administration, in good stuck to his guns, trying to control inflation even part because C. Douglas Dillon, the secretary of if it meant erring on the side of overtightening the the Treasury and a Rockefeller Republican, sided economy. Our estimated γ captures this atti tude on many occasions with Martin against Heller. 12 Π, t with an increase from around 1965 to around 1968. But the changing composition of the Board of But by the summer of 1968, Martin gave in Governors and the arrival of Johnson, with his to an easing of monetary policy after the tax sur - expansionary fiscal programs, the escalation of charge was passed by Congress. As reported by the Vietnam War, and the departure of Dillon from Hetzel (2008), at the time, the FOMC was divided the Treasury Department, changed the weights of power. into two camps: members more concerned about While the effects of the expansion of federal inflation (such as Al Hayes, the president of the spending in the second half of the 1960s often Federal Reserve Bank of New York) and mem bers 14 play a central role in the narrative of the start of more concerned about output growth (Brimmer, 15 the Great Inflation, the evolution of the Board of Maisel, and Mitchell, all three appointees of Governors has received less attention. Heller Kennedy and Johnson). Martin, always a seeker realized that, by carefully selecting the gover nors, of consensus, was growlingly incapable of carry - he could shape monetary policy without the ing the day. 16 Perhaps Martin felt that the politi - need to ease Martin out. This was an inspired cal climate had moved away from a commitment observation, since up to that moment, the gover - nors who served under the chairman had played 13 Oral history interview with Martin, Lyndon B. Johnson Library an extremely small role in monetary policy and (quoted by Bremner, 2004, p. 191). the previous administrations had, consequently, 14 Brimmer is also the first African American to have served as a governor and was, for a while, a faculty member at the University shown little interest in their selection. The strat - of Pennsylvania.

15 Sherman Maisel was a member of the Board of Governors between 11 The message of the paper is, however, much more subtle than 1965 and 1972. Maisel, a professor at the Haas School of Business lay ing down a simple textbook Phillips curve. As Samuelson and at the University of California–Berkeley, has the honor of being the Solow (1960) also say in the next page of their article (p. 193), “All first academic economist appointed as a governor after Adolph of our discussion has been phrased in short-run terms, dealing Miller, one of the original governors in 1914. As he explained in with what might happen in the next few years. It would be wrong, his book, Managing the Dollar (one of the first inside looks at the though, to think that our Figure 2 menu that relates obtainable price Fed and still a fascinating read today), Maisel was also a strong and unemployment behavior will maintain its shape in the longer believer in the Phillips curve: “There is a trade-off between idle run. What we do in a policy way during the next few years might men and a more stable value of the dollar. A conscious decision cause it to shift in a definite way.” must be made as to how much unemployment and loss of output is acceptable in order to get smaller price rises” (Maisel, 1973, 12 In particular, Dillon’s support for Martin’s reappointment for a p. 285). Maisel’s academic and Keynesian background merged in new term in 1963 was pivotal. Hetzel (2008, p. 69) suggests that his sponsoring of the MPS model mentioned in Section 2. President Kennedy often sided with Dillon and Martin over Heller to avoid a gold crisis on top of the problems with the Soviet Union 16 On one occasion, Maisel felt strongly enough to call a press con - over Cuba and Berlin. ference to explain his dissenting vote in favor of more expansion.

16 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

17 to fight inflation. Or perhaps he was just Our smoothed estimate of γ Π, t in Figure 2 exhausted after many years running the Fed (at responds to this behavior of the Fed by quickly the last meeting of the FOMC in which he partic i- dropping during the same period. This indicates pated, he expressed feelings of failure for not that the actual reduction in the federal funds rate having controlled inflation). No matter what the was much more aggressive than the reduction exact reason was, monetary policy eased drasti - suggested by the (important) fall in output growth cally in comparison with what was being called and the (moderate) fall in inflation. Furthermore, the likelihood function accounts for the persist ent for by the Taylor rule with a γ Π, t above 1. Thus, fall in the real interest rate with a persistent fall our estimated γ Π, t starts to plunge in the spring of 1968, reflecting that the increases in the fed eral in γ Π, t. funds rate passed at the end of 1968 and in 1969 Burns did little over the next few years to were, according to our estimated Taylor rule, not return γ Π, t to higher values. Even if the federal aggressive enough given the state of the econ omy. funds rate had started to grow by the end of 1971 The genie of the Great Inflation was out of the (after the 90-day price controls announced on bottle. August 15 of that year as part of Nixon’s New Economic Policy) and reached new highs in 1973 and 1974, it barely kept up with inflation. The 6.2 The Burns-Miller Era: Monetary real interest rate was not above our benchmark Policy in the Time of Turbulence value of 2 percent until the second quarter of 1976. Arthur F. Burns started his term as chairman Later, in 1977, the federal funds rate was only of the Fed on February 1, 1970. A professor of raised cautiously, despite the evidence of strong economics at Columbia University and the presi - output growth after the 1973-75 recession and dent of the National Bureau of Economic Research that inflation remained relatively high. between 1957 and 1967, Burns was the first aca - Our econometric results come about because demic economist to hold the chairmanship. All the Taylor rule does not care about the level of the previous nine chairmen had been bankers or the interest rate in itself, but by how much infla - lawyers. However, any hope that his economics tion deviates from Π. If γ Π, t > 1, the increases in education would make him take an aggressive the federal funds rate are bigger than the stand against the inflation brewing during the increases in inflation. This is not what happened last years of Martin’s tenure quickly disap peared. during Burns’s tenure: The real interest rate was The federal funds rate fell from an average of above the cutoff of 2 percent that we proposed 8.02 percent during 1970:Q1 to 4.12 percent by before only in three quarters: his first two quar - 1970:Q4. The justification for those reductions ters as chairman (1970:Q2 and 1970:Q3) and in was the need to jump-start the economy, which 1976:Q2. This observation, by itself, should be was stacked in the middle of the first recession sufficient proof of the stand of monetary policy in nearly a decade, since December 1969. But during the period. 18 since inflation stayed at 4.55 percent by the end Burns’s successor, William Miller, did not of 1970, the reduction in the nominal rate meant have time to retract these policies in the brief that real interest rates sank into the negative interlude of his tenure, from March 8, 1978, to region. August 6, 1979. But he also did not have either the capability, since his only experience in the 17 Meltzer (2010, p. 549) points out that Martin and the other Board conduct of monetary policy was serving as a members might have been worried by Johnson’s appointment, at the suggestion of Arthur Okun (the chairman of the Council of Economic Advisors at the time), of a task force to review changes 18 A memorandum prepared at the end of December 1997 by two of in the Federal Reserve System. That message only became rein - Carter’s advisers reveals the climate of the time, proposing not to forced with the arrival of a new administration in 1969, given reappoint Chairman Burns for a third term because he was more Richard Nixon’s obsession with keeping unemployment as low as concerned with inflation than unemployment (memo for the presi - possible. (Nixon was convinced that he had lost the 1960 presi den - dent on the role of the Federal Reserve, Box 16, R.K. Lipshitz Files, tial election to a combination of vote fraud and tight monetary Carter Library, December 10, 1977, pp. 1-2; cited by Meltzer, 2010, policy.) p. 922).

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 17 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez director of the Federal Reserve Bank of Boston, Board back then, Burns did not believe any the ory or the desire, since he had little faith in restric tive of the economy—whether Keynesian or mone - monetary policy’s ability to lower inflation. 19 tarist—could account for the business cycle; he

Thus, our estimated γ Π, t remains low during that dismissed the relation between the stock of money time. 20 and the price level; and he was unwilling or Burns was subject to strong pressure from unable to make a persuasive case against infla tion Nixon. 21 His margin of maneuver was also lim ited to the nation and to the FOMC. 23 by the views among many leading economists In addition, Burns had a sympathetic atti tude that overestimated the costs of disinflation and toward price and wage controls. For instance, he who were in any case skeptical of monetary pol - testified to Congress on February 7, 1973: icy. 22 But his own convictions leaned in the same direction. According to the recollections of [T]here is a need for legislation permitting some direct controls over wages and prices...The Stephen H. Axilrod, a senior staff member at the structure of our economy—in particular, the power of many corporations and trade unions 19 Miller stated, “Our attempts to restrain inflation by using conven - to exact rewards that exceed what could be tional stabilization techniques have been less than satisfactory. Three years of high unemployment and underutilized capital stock achieved under conditions of active competi - have been costly in terms both of lost production and of the denial tion—does expose us to upward pressure on to many of the dignity that comes from holding a productive job. costs and prices that may be cumulative and Yet, despite this period of substantial slack in the economy, we still have a serious inflation problem” (Board of Governors, 1978, self-reinforcing (cited by Hetzel, 2008, p. 79). p. 193; quoted by Romer and Romer, 2004, p. 140). He reiterated that view in a letter to the pres i- 20 The situation with Miller reached the surrealistic point when, as narrated by Kettl (1986), Charles Schultze, the chairman of the dent on June 1, 1973, in which he proposed to Council of Economic Advisors, and Michael Blumenthal, the reintroduce mandatory price controls for large Treasury secretary, were leaking information to the press to pres sure firms. 24 In his view, controls could break the cost- Miller to tighten monetary policy. push spiral of the economy and the inflationary 21 Perhaps the clearest documented moment is the meeting between Nixon and Burns on October 23, 1969, right after Burns’s nomina - pressures triggered by the social unrest of the late tion, as narrated by John Ehrlichman (1982, pp. 248-49): “I know 1960s and be a more effective instrument than there’s the myth of the autonomous Fed...Nixon barked a quick laugh…and when you go up for confirmation some Senator may open market operations, which could be quite ask you about your friendship with the President. Appearances costly in terms of employment and financial dis - are going to be important, so you can call Ehrlichman to get mes - turbances. 25 In fact, many members of the FOMC sages to me, and he’ll call you.” The White House continued its pressure on Burns by many different methods, from constant con - believed that the introduction of price and wage versations to leaks to the press (falsely) accusing Burns of request - controls in different phases between 1971 and ing a large wage increase. These, and many other histories, are collected in a fascinating article by Abrams (2006). 1973 had not only eased the need for monetary

22 Three examples. First, Franco Modigliani testified before the U.S. tightening, but also positively suggested that mone - Congress on July 20, 1971: “[Y]ou have to recognize that prices are tary policy should not impose further restraint presently rising, and no measure we can take short of creating mas - 26 sive unemployment is going to make the rate of change of prices on the economy. More interestingly, if price and substantially below 4 percent.” Second, Otto Eckstein, the builder wage controls were an argument for loose mone - of one of the large macroeconometric models at the time, the DRI tary policy, their easing was also an argument U.S. model, argued that it was not the Fed’s job to solve structural inflation. Third, James Tobin (1974): “For the rest of us, the tor ment - for expansionary policy, or as Governor Charles ing difficulty is that the economy shows inflationary bias even Partee put it during the FOMC meeting of when there is significant involuntary unemployment. The bias is in some sense a structural defect of the economy and society… January 11, 1973, the lifting of controls “might Chronic and accelerating inflation is then a symptom of a deeper necessitate a somewhat faster rate of monetary social disorder, of which involuntary unemployment is an alter - native symptom. Political economists may differ about whether it is better to face the social conflicts squarely or to let inflation 24 Burns papers, B_N1, June 1, 1973, as cited by Meltzer (2010, p. 787). obscure them and muddle through. I can understand why anyone who prefers the first alternative would be working for structural 25 At the time, many financial institutions were subject to ceiling reform, for a new social contract. I cannot understand why he rates on deposits, which could have made them bankrupt in the would believe that the job can be done by monetary policy. Within case of a fast tightening of monetary policy. limits, the Federal Reserve can shift from one symptom to the other. But it cannot cure the disease.” The examples are quoted by Hetzel 26 Maisel’s diary entry for August 25, 1971; cited by Meltzer, 2010, (2008, pp. 86, 89, and 128). p. 790.

18 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez growth to finance the desired growth in real out - increase wages, which pushes up the marginal put under conditions of greater cost-push infla tion cost and, therefore, inflation. Moreover, FGR show than would have prevailed with tighter con trols” that, if volatility had stayed at historical levels, (cited by Meltzer, 2010, p. 815). even with negative innovations, inflation would Burns’s 1979 Per Jacobsson lecture is a reveal - have been much lower and the big peak of 1973 ing summary of Burns’s own views on the ori gins avoided. and development of inflation. He blamed the However, these negative shocks should not growing demands of different social groups dur ing make us forget that, according to our model, if the late 1960s and early 1970s and the federal monetary policy had engineered higher real government’s willingness to concede to them as interest rates during those years, the history of the real culprit behind inflation. Moreover, he inflation could have been different. In FGR we felt that the Fed could not really stop the infla tion - calculate that, had monetary policy behaved ary wave: If the Federal Reserve then sought to under Burns and Miller as it did under Volcker, create a monetary environment that fell seri ously inflation would have been 4.36 percent on aver age, short of accommodating the upward pressures instead of the observed 6.23 percent. The experi - on prices that were being released or reinforced ence of Germany or Switzerland, which had much by governmental action, severe difficulties could lower inflation than the United States during the be quickly produced in the economy. Not only same time, suggests that this was possible. After that, the Federal Reserve would be frustrating all, the peak of inflation in Germany was in 1971, the will of Congress to which it was responsible. well before any of the oil shocks. And in neither But beyond Burns’s own defeatist attitude of these two European countries do we observe toward inflation, he was a most unfortunate chair - statements such as that by Governor Sheehan on man. He was in charge during a period of high the January 22, 1974, FOMC meeting: “[T]he turbulence and negative shocks, not only the 1973 Committee had no choice but to validate the rise oil shock, but also poor crops in the United States in prices if it wished to avoid compounding the and the Soviet Union. Our model estimates large recession” (Hetzel, 2008, p. 93). Thus, our reading of monetary policy during and volatile intertemporal shocks, dt, and labor the Burns years through the lens of our model supply shocks, ϕt, during his tenure (see FGR for a plot of these shocks). Examples of intertempo ral emphasizes the confluence of two phenomena: shocks include the final breakdown of the Bretton an accommodating position with respect to infla - Woods Agreement, fiscal policy during the 1973- tion and large and volatile shocks that compli - 75 recession (with a temporary tax cut signed in cated the implementation of policy. There is ample March 1975 and increases in discretionary spend - evidence in the historical record to support this ing), and Nixon’s price and wage controls (which view. This was, indeed, monetary policy in the most likely distorted intratemporal allocations). time of turbulence. Examples of labor supply shocks include the his - torically high level of strikes in American indus - 6.3 The Volcker Era: High Noon try during the early 1970s. (A major issue in the In his 1979 Per Jacobsson lecture cited ear lier, Republican primary of 1976 between Ford and Burns had concluded: “It is illusory to expect Reagan was picketing rules for striking workers, central banks to put an end to the inflation that a policy issue most unlikely to grab many voters’ now afflicts the industrial democracies.” Paul attention nowadays.) Volcker begged to differ. He had been president Both types of shocks complicated monetary of the Federal Reserve Bank of New York since policy. Large positive intertemporal shocks August 1975 and, from that position, a vocal foe increase aggregate demand. In our model, this of inflation. In particular, during his years as a translates partly into higher output and partly member of the FOMC, Volcker expressed con cern into higher inflation. Positive labor supply shocks that the Fed was consistently underpredicting

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 19 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez inflation and that, therefore, monetary policy until October 6, 1979. Additionally, Goodfriend was more expansionary than conventionally and King (2005) have argued that Volcker required understood (Meltzer, 2010, p. 942). 27 some time before asserting his control over the In the summer of 1979, Jimmy Carter moved FOMC. For instance, in the Board meeting of Miller to the Treasury Department. Then, he September 18, 1979, Volcker did obtain a rise in offered Volcker the chairmanship of the Board of the discount rate, but only with three dissenting Governors. Volcker did not hesitate to take it, but votes. As we argued in Section 2, all of these not before warning the president “of the need observations suggest that modeling the evolution for tighter money—tighter than Bill Miller had of monetary policy as a smooth change may be wanted” (Volcker and Gyothen, 1992, p. 164) more appropriate than assuming a pure break. and the Senate in his confirmation hearings that Regardless of the exact timing of changes in “the only sound foundation for the continuing monetary policy, the evidence in Figure 2 is over - growth and prosperity of the American economy whelming: On or about August 1979, the charac ter is much greater price stability” (U.S. Senate, 1979, of monetary policy changed. The federal funds p. 16; quoted by Romer and Romer, 2004, p. 156). rate jumped to new levels, with the first signifi - Deep changes were coming and the main deci sion - cant long-lasting increase in the real interest rate makers were aware of them. in many years. Real interest rates would remain We should be careful not to attribute all of high for the remainder of the decade of the 1980s, the sharp break in monetary policy to Volcker’s partly reflecting high federal fund rates and partly appointment. In 1975, the House passed Concur - reflecting the deeply rooted expectations of infla - rent Resolution 133, the brainchild of Karl Brunner tion among the agents. In any case, the response (Weintraub, 1977). This resolution, which asked of monetary policy to inflation, γ Π, t, was consis - the Fed to report to the House Banking Committee tently high during the whole of Volcker’s years. on objectives and plans with respect to the ranges An important question is the extent to which of growth or diminution of monetary and credit the formalism of the Taylor rule can capture the aggregates in the upcoming twelve months, was way in which monetary policy was conducted a first victory for monetarism. Although the reso - at the time, when money growth targeting and lution probably did little by itself, it was a sign reserve management were explicitly tried (what that times were changing. Congress acted again Volcker called “practical monetarism”). We are with the Full Employment and Balanced Growth not overly concerned about this aspect of the data Act of 1978, which required the Fed to report because, in our DSGE model, there is a mapping monetary aggregates in its reports to Congress. In between money targeting and the Taylor rule April 1978, the federal funds rate started grow ing (Woodford, 2003). Thus, as long as we are care ful quickly, from a monthly average of 6.9 percent to to interpret the monetary policy shocks during the 10 percent by the end of the year. This reflected period (which we estimate were, indeed, larger a growing consensus on the FOMC (still with than in other parts of the sample), our exercise many dissenting voices) regarding the need for should be relatively robust to this consideration. 28 lower inflation. Figure 2 shows the start of an A much more challenging task could be to build increase in γ around that time. At the same time, Π, t a DSGE model with a richer set of monetary pol - the new procedures for monetary policy that tar - geted money growth rates and reserves instead 28 This begets the question of why Volcker spent so much effort on of the federal funds rate were not announced switching the operating procedure of the Fed between 1979 and 1982. Volcker himself ventures that it was easier to sell a restric tive monetary policy in terms of money growth rates than in terms of 27 This position links to an important point made by Orphanides interest rates: “More focus on the money supply also would be a (2002): Monetary policy decisions are implemented using real- way of telling the public that we meant business. People don’t need time data, a point that our model blissfully ignores. In turbulent an advanced course in economics to understand that inflation has times such as the 1970s, this makes steering the ship of policy tar - something to do with too much money” (Volcker and Gyohten, gets exceedingly difficult. 1992, pp. 167-68).

20 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez icy rules and switches between them. However, of the intratemporal preference shifter did not at the moment, this goal seems infeasible. 29 fall until later in his term. In FGR, we build a The impressions of participants in the mone - counterfactual in which Volcker is faced with tary policy process reinforced the message of the same structural shocks he faced in real life, Figure 2. For instance, Axilrod (2009, p. 91) states: but with the historical average volatility. In this During Paul Volcker’s eight-year tenure as counterfactual history, inflation falls to negative chairman of the Fed...policy changed dramat i- values by the end of 1983, instead of still hover - cally. He was responsible for a major transfor - ing around 3 to 4 percent. It was a tough policy mation—akin to a paradigm shift—that was in a difficult time. However, despite these mis for - intended to greatly reduce inflation, keep it tunes and heavy inheritance from the past, our under control, and thereby restore the Fed’s model tells us that monetary policy conquered badly damaged reputation. Furthermore, it was the Great Inflation. The Great Moderation would almost solely because of Volcker that this par - ticular innovation was put in place—one of the have to wait for better shocks. few instances in my opinion where a dra matic We started this subsection with Burns’s own shift in policy approach could be attributed words in the 1979 Per Jacobsson lecture. In 1989, to a particular person’s presence rather than Volcker was invited to give the same lecture. What mainly or just to circumstances. a difference a decade can make! While Burns Volcker himself was very explicit about his was sad and pessimistic (his lecture was entitled views 30 : “The Anguish of Central Banking”), Volcker was happy and confident (his lecture was entitled [M]y basic philosophy is over time we have no “The Triumph of Central Banking?”). Inflation choice but to deal with the inflationary situa - tions because over time inflation and unem - had been defeated and he warned that “our col lec - ployment go together...Isn’t that the lesson of tive experience strongly emphasizes the impor - the 1970s? We sat around [for] years thinking tance of dealing with inflation at an early stage” we could play off a choice between one or the (Volcker, 1990, p. 14). other...It had some reality when everybody thought processes were going to be stable...So 6.4 The Greenspan Era: Speaking Like in a very fundamental sense, I don’t think we have the choice. a Hawk and Walking Like a Dove These are the colorful words with which In fact, Volcker’s views put him in the rather Laurence Meyer (2004, p. 83) summarizes unusual position of being outvoted on February 24, 1986. In that meeting, a majority of four mem bers Greenspan’s behavior during Meyer’s time as a of the Board voted against Volcker and two other governor (June 1996 to January 2002). Once and dissenting members to lower the discount rate again, 50 basis points. [Greenspan] seemed to fall into a pattern: The At the same time, and according to our model, Chairman would ask for no change in the funds Volcker was also an unlucky chairman. The econ - rate, suggest that the time was approaching omy still suffered from large and negative shocks for action, and indicate that there was a high during his tenure, since the level and volatility probability of a move at the next meeting. Then at the next meeting, he would explain that the

29 The impact of the credit controls imposed by the Carter adminis - data did not yet provide a credible basis for tration starting on March 14, 1980, is more difficult to gauge. Inter - tightening, and in any case, that the markets estingly, we estimate a large negative innovation to the intratemporal didn’t expect a move. However, he would con - preference shifter at that point in time, a likely reflection of the distortions of the controls in the intertemporal choices of house - clude that he expected the Committee would holds (see the historical description in Shreft, 1990). be forced to move at the next meeting. 30 Volcker papers, Federal Reserve Bank of New York, speech at the Meyer means these words in a positive way. In National Press Club, Box 97657, January 2, 1980; quoted by Meltzer, 2010, p. 1034. his opinion, Greenspan discovered before he did

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 21 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez that the economy was being hit during the sec ond would have fallen by only 13 percent (instead of half of the 1990s by an unusual sequence of pos i- 60 percent in the data), the SD of output growth tive shocks and directed monetary policy to take would have fallen by 16 percent (instead of 46 advantage of them. percent in the data), and the SD of the federal funds We quote Meyer because it illustrates that rate would have fallen by 35 percent (instead of Greenspan showed from the start that he knew 39 percent in the data). That is, the moderation how to respond to changing circumstances. He in inflation fluctuations would have been only was appointed on August 11, 1987. In his confir - one-fifth as large as in the data (and the counter - mation hearings, he clearly reaffirmed the need to factual mean would have actually been higher fight inflation. 31 But after just a couple of months, than in the data) and the moderation for output on October 19, 1987, he reacted to the big crash growth’s SD only one-third. of the stock market by declaring the Fed’s dispo - We can push the argument even further. In sition to serve as a source of liquidity, even if, in FGR we build the counterfactual in which the the short run, this could complicate the control average γ Π, t during the Greenspan years is plugged of inflation. into the model at the time of Burns’s appoint ment. Later, in early 1989, the federal funds rate Then, we keep γ Π, t at that level and hit the model started to fall, despite the fact that inflation with exactly the same shocks that we backed out remained at around 6 percent until the end of from our estimation. This exercise is logically

1990. As shown in Figure 2, our estimate of γ Π, t coherent, since we are working with a DSGE picks up this fall by dropping itself. Moreover, it model and, therefore, the structural and volatil ity dropped fast. We estimate that γ Π, t was soon below shocks are invariant to this class of interven tions. 1, back to the levels of Burns-Miller (although, for We compute that the average monetary policy a while, there is quite a bit of uncertainty in our during Greenspan’s years would not have made estimate). The parameter stayed there for the rest much of a difference in the 1970s. If anything, of Greenspan’s tenure. The reason for this esti - inflation would have been even slightly higher mated low level of γ Π, t is that the real interest rate (6.83 percent in the counterfactual instead of 6.23 also started to fall rather quickly. At the same time, percent in the data). This finding contrasts with a remarkable sequence of good shocks delivered our counterfactual in which Volcker is moved to rapid output growth and low inflation. the Burns-Miller era. In this counterfactual, infla - In fact, in FGR we find that all of the shocks tion would have been only 4.36 percent. To sum - went right for monetary policy during the 1990s. marize, our reading of monetary policy during the A large string of positive and stable investment- Greenspan years is that it was not too different specific technological shocks delivered fast pro - from the policy in the Burns-Miller era; it just ductivity growth, a falling intertemporal shifter faced much better shocks. lowered demand pressures, and labor supply Is this result credible? First, it is clear that it shocks pressured wages downward—and, with is not a pure artifact of our model. A similar them, marginal costs. This fantastic concatena tion result is found in Sims and Zha (2006). These of shocks accounted for the bulk of the Great authors, using structural vector autoregressions Moderation. In FGR, we calculate that without with Markov switching, which imposes many changes in volatility, the Great Moderation would fewer cross-equation restrictions than our analy - have been much smaller. The SD of inflation sis, do not find much evidence of differences in monetary policy across time (actually, Sims and 31 Greenspan stated in his confirmation hearings: “[W]e allowed our Zha’s position is even stronger than ours, since system to take on inflationary biases which threw us into such a they do find that monetary policy was different structural imbalance that, in order to preserve the integrity of the system, the Federal Reserve had to do what it did. Had it not acted even under Volcker). Second, there are hints in in the way which it did at that time, the consequences would have the data that lead us to believe that the results been far worse than what subsequently happened” (U.S. Senate, 1987, p. 35; quoted by Romer and Romer, 2004, p. 158). make sense. At the start of the 1994 inflation scare,

22 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez when there were no signs of the “new economy” the behavior of policy in the aftermath of the Long anywhere to be seen, Greenspan argued 32 : Term Capital Management fiasco or in the exit from the 2001 recession. But we feel the point You know, I rarely feel strongly about an issue, and I very rarely sort of press this Committee. has been made. We believe that our estimates are But let me tell you something about what’s right: Monetary policy in the Greenspan years gnawing at me here. I am very sympathetic was similar to monetary policy under Burns and with the view that we’ve got to move and that Miller. Instead, time-varying structural shocks we’re going to have an extended period of were the mechanism that played a key role in moves, assuming the changes that are going on the Great Moderation and the low inflation of now continue in the direction of strength. It is 1987-2007. very unlikely that the recent rate of economic growth will not simmer down largely because some developments involved in this particu lar period are clearly one-shot factors—namely, 7. WHAT ARE WE MISSING? the very dramatic increase in residential con - What is our model missing that is really impor - struction and the big increase in motor vehi cle tant? The answer will tell us much about where sales. Essentially the two of those have added we want to go in terms of research and where we one-shot elements to growth. In the context of need to be careful in our reading of monetary his - a saving rate that is not high, the probability is tory. Of all the potential problems of our specifi - in the direction of this expansion slowing from cation, we are particularly concerned about the its recent pace, which at the moment is well following. over 4 percent and, adjusting for weather First, households and firms in the model effects, may be running over 5 percent. This is observe the changes in the coefficients γ and γ not sustainable growth, and it has nothing to Πt yt when they occur. A more plausible scenario would do with monetary policy. In other words, it will come down. And the way a 3 percent growth involve filtering in real time by the agents who feels, if I may put it that way, is a lot different need to learn the stand of the monetary authority 33 from the way the expansion feels now. from observed decisions. A similar argument I would be very concerned if this Committee can be made for the values of the SDs of all the went 50 basis points now because I don’t think other shocks in the economy. Unfortunately, the markets expect it...I’ve been in the eco nomic introducing learning suffers from two practical forecasting business since 1948, and I’ve been difficulties: It is not obvious what is the best way on Wall Street since 1948, and I am telling you to model learning about monetary policy, espe - I have a pain in the pit of my stomach, which cially in a nonlinear environment such as ours in the past I’ve been very successful in allud ing where least-square rules may not work properly. to. I am telling you—and I’ve seen these mar - And it would make the computation of the model kets—this is not the time to do this. I think there nearly infeasible. will be a time; and if the staff’s forecast is right, Second, we assume that monetary policy we can get to 150 basis points pretty easily. changes are independent of the events in the 1 We can do it with a couple of /2 point jumps economy. However, many channels make this later when the markets are in the position to assumption untenable. For instance, each admin - know what we’re doing and there’s continu ity. istration searches for governors of the Board who I really request that we not do this. I do request that we be willing to move again fairly soon, conform with its views on the economy (after all, and maybe in larger increments; that depends on how things are evolving. 33 The difficulties in observing monetary policy changes can be illustrated by Axilrod’s description of a lunch with Arthur Burns We construe this statement as revealing a low shortly after the announcement of Volcker’s new policy. According to Axilrod (2009, p. 100), Burns stated: “You are not really going γ Πt. We could present similar evidence regarding to be doing anything different from what we were doing. If an insider like Burns had difficulties in filtering Volcker’s behavior, it is hard to conclude anything but that the average agents in the 32 Board of Governors FOMC Transcripts, February 3-4, 1994, p. 55. economy had difficulties as well.”

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 23 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez this is what a democracy is supposed to be about). the early 1970s. Later, Volcker’s last years at the We saw how Heller discovered that an adminis - Fed were colored by the Plaza and Louvre Accords tration could select governors to twist the FOMC and the attempts to manage the exchange rate toward its policy priorities. This is a tradition between the U.S. dollar and the Japanese yen. that has continued. Meyer (2004, p. 17) describes Finally, our model ignores fiscal policy. The the process for his own appointment as one clearly experience of the 1960s, in which there was an guided by the desire of the Clinton administra tion explicit attempt at coordinating fiscal and mone - to make monetary policy more accommodative tary policies, and the changes in long-run inter est and growth-oriented. As long as the party in power rates possibly triggered by the fiscal consolida - is a function of the state of the economy, the com - tions of the 1990s indicate that the interaction position of the FOMC will clearly be endoge nous. between fiscal and monetary policies deserves Similarly, changes in public perception of the much more attention, a point repeatedly made dangers of inflation certainly weighed heavily by Chris Sims (for example in Sims, 2009). on Carter when he appointed Volcker to lead the Fed in 1979. Third, and related to our two previous points, evolving beliefs about monetary policy might be 8. CONCLUDING REMARKS endogenous to the developments of events and The title of this paper is not only a tribute to lead to self-confirming equilibria. This is a point Friedman and Schwartz’s (1971) opus magnum, emphasized by Cho, Williams, and Sargent (2002) but also a statement of the limitations of our inves - and Sargent (2008). tigation. Neither the space allocated to us 34 nor Fourth, our technological drifts are constant our own abilities allow us to get even close to over time. The literature on long-run risk has high - Friedman and Schwartz’s achievements. We have lighted the importance of slow-moving compo - tried to demonstrate, only, that the use of mod ern nents in growth trends (Bansal and Yaron, 2004). equilibrium theory and econometric methods It may be relevant to judge monetary policy to esti - allows us to read the monetary policy history of mate a model in which we have these slow-mov ing the United States since 1959 in ways that we find components, since the productivity slowdown fruitful. We proposed and estimated a DSGE model of the 1970s and the productivity acceleration of with stochastic volatility and parameter drifting. the late 1990s are bound to be reflected in our The model gave us a clear punch line: First, there assessment of the stance of monetary policy dur - is ample evidence of both strong changes in the ing those years. This links us back to some of the volatility of the structural shocks that hit the econ - concerns expressed by Orphanides (2002). At the omy and changes in monetary policy. The changes same time and nearly by definition, there is very in volatility accounted for most of the Great little information in the data about this compo nent. Moderation. The changes in monetary policy Fifth, our model is a closed economy. How - mattered for the rise and conquest of the Great ever, the considerations regarding exchange rates have often played an important role in monetary Inflation. Inflation stayed low during the next policymaking. For instance, during the late 1960s, decades in large part due to good shocks. When the United States fought an increasingly desper ate we go to the historical record and use the results battle to keep the Bretton Woods Agreement of our estimation to read and assess the docu men - alive, which included the Fed administering a tary evidence, we find ample confirmation, in our program to voluntarily reduce the amount of opinion, that the model, despite all its limita tions, funds that American banks could lend abroad is teaching us important lessons. (Meltzer, 2010, p. 695) and purchasing long-term Treasury bonds to help the British pound stabi lize 34 For an only slightly longer period than ours, Meltzer (2010) requires after its 1967 devaluation. The end of Bretton 1,300 pages to cover the details of the history of monetary policy in the United States, including the evolution of operational pro - Woods also deeply influenced policymakers in cedures that we have not even mentioned.

24 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

As we argued in the previous section, we leave much unsaid. Hopefully, the results in this paper will be enticing enough for other researchers to continue a close exploration of recent mone tary policy history with the tools of modern dynamic macroeconomics.

REFERENCES Abrams, Burton A. “How Richard Nixon Pressured Arthur Burns: Evidence from the Nixon Tapes.” Journal of Economic Perspectives , Fall 2006, 20 (4), pp. 177-88.

Aruoba, S. Boragan and Schorfheide, Frank. “Sticky Prices versus Monetary Frictions: An Estimation of Policy Trade-offs.” American Economic Journal: Macroeconomics (forthcoming).

Aruoba, S. Boragon; Fernández-Villaverde, Jesus and Rubio-Ramírez, Juan F. “Comparing Solution Methods for Dynamic Equilibrium Economies.” Journal of Economic Dynamics and Control , December 2006, 30 (12), pp. 2477-508.

Axilrod, Stephen H. Inside the Fed: Monetary Policy and Its Management, Martin through Greenspan to Bernanke . Cambridge, MA: MIT Press, 2009.

Bansal, Ravi and Yaron, Amir. “Risks for the Long Run: A Potential Resolution of Asset Pricing Puzzles.” Journal of Finance , August 2004, 59 (4), pp. 1481-509.

Board of Governors of the Federal Reserve System. “Statement by G. William Miller, Chairman, Board of Governors of the Federal Reserve System, before the Committee on the Budget, U.S. Senate, March 15, 1978,” in Federal Reserve Bulletin , March 1978, 64 (3), pp. 190-94. http://fraser.stlouisfed.org/publications/frb/1978/download/60182/frb_031978.pdf.

Board of Governors of the Federal Reserve System. Meeting of the Federal Open Market Committee, February 3-4, 1994 (FOMC Transcripts); www.federalreserve.gov/monetarypolicy/files/FOMC19940204meeting.pdf.

Brayton, Flint; Levin, Andrew; Lyon, Ralph and Williams John C. “The Evolution of Macro Models at the Federal Reserve Board.” Carnegie-Rochester Conference Series on Public Policy , December 1997, 47 (1), pp. 43-81.

Bremner, Robert P. Chairman of the Fed: William McChesney Martin Jr. and the Creation of the American Financial System . New Haven, CT: Yale University Press, 2004.

Burns, Arthur F. “The Anguish of Central Banking.” The 1979 Per Jacobbson Lecture, Belgrade, Yugoslavia, September 30, 1979; www.perjacobsson.org/lectures/1979.pdf.

Christiano, Lawrence; Eichenbaum, Martin and Evans, Charles L. “Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy.” Journal of Political Economy , June 2005, 113 (1), pp. 1-45.

Cho, In-Koo; Williams, Noah and Sargent, Thomas J. “Escaping Nash Inflation.” Review of Economic Studies , January 2002, 69 (1), pp. 1-40.

Clarida, Richard; Galí, Jordi and Gertler, Mark. “Monetary Policy Rules and Macroeconomic Stability: Evidence and Some Theory.” Quarterly Journal of Economics , February 2000, 115 (1), pp. 147-80.

Cogley, Timothy and Sargent, Thomas J. “Evolving Post-World War II U.S. Inflation Dynamics,” in Ben S. Bernanke and Kenneth Rogoff, eds., NBER Macroeconomics Annual 2001 . Volume 16. Cambridge, MA: MIT Press, 2002, pp. 331-88.

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 25 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Cogley, Timothy and Sargent, Thomas J. “Drifts and Volatilities: Monetary Policies and Outcomes in the Post WWII U.S.” Review of Economic Dynamics , April 2005, 8(2), pp. 262-302. Dotsey, Michael; King, Robert G. and Wolman, Alexander L. “State-Dependent Pricing and the General Equilibrium Dynamics of Money and Output.” Quarterly Journal of Economics , May 1999, 114 (2), pp. 655-90. Ehrlichman, John. Witness to Power: The Nixon Years . New York: Simon and Schuster, 1982. Fernández-Villaverde, Jesus and Rubio-Ramírez, Juan F. “Estimating Macroeconomic Models: A Likelihood Approach.” Review of Economic Studies , October 2007, 74 (4), pp. 1059-87. Fernández-Villaverde, Jesus and Rubio-Ramírez, Juan F. “How Structural Are Structural Parameters?” in Daron Acemoglu; Kenneth Rogoff and Michael Woodford, eds., NBER Macroeconomics Annual 2007 . Volume 22. Chicago: University of Chicago Press, 2008, pp. 83-137. Fernández-Villaverde, Jesus; Guerrón-Quintana, Pablo and Rubio-Ramírez, Juan F. “Fortune or Virtue: Time- Variant Volatilities Versus Parameter Drifting in U.S. Data.” NBER Working Paper No. 15928, National Bureau of Economic Research, April 2010; www.nber.org/papers/w15928. Fisher, Jonas. “The Dynamic Effects of Neutral and Investment-Specific Technology Shocks.” Journal of Political Economy , June 2006, 114 (3), pp. 413-52. Friedman, Milton and Schwartz, Anna J. A Monetary History of the United States, 1867-1960 . Princeton, NJ: Princeton University Press, 1971. Goodfriend, M. and King, Robert. “The Incredible Volcker Disinflation.” Journal of Monetary Economics , July 2005, 52 (5), pp. 981-1015. Greenwood, Jeremy; Herkowitz, Zvi and Krusell, Per. “Long-Run Implications of Investment-Specific Technological Change.” American Economic Review , June 1997, 87 (3), pp. 342-62. Hall, Robert E. “Macroeconomic Fluctuations and the Allocation of Time.” Journal of Labor Economics , January 1997, 15 (1 Part 1), pp. S223-S250. Hetzel, Robert L. The Monetary Policy of the Federal Reserve: A History . New York: Cambridge University Press, 2008. Hurwicz, Leonid. “On the Structural Form of Interdependent Systems,” in Ernest Nagel, Patrick Suppes, and Alfred Tarski, eds., Logic, Methodology and Philosophy of Science: Proceedings of the 1960 International Congress . Stanford, CA: Stanford University Press, 1962, pp. 232-39. Jaimovich, Nir and Siu, Henry E. “The Young, the Old, and the Restless: Demographics and Business Cycle Volatility.” American Economic Review , June 2009, 99 (3), pp. 804-26. Judd, Kenneth L. Numerical Methods in Economics . Cambridge, MA: MIT Press, 1998. Justiniano, Alejandro and Primiceri, Giorgio. “The Time Varying Volatility of Macroeconomic Fluctuations.” American Economic Review , June 2008, 98 (3), pp. 604-41. Kettl, Donald. Leadership at the Fed . New Haven, CT: Yale University Press, 1986. Kim, Chang-Jin and Nelson, Charles R. “Has the U.S. Economy Become More Stable? A Bayesian Approach Based on a Markov-Switching Model of the Business Cycle.” Review of Economics and Statistics , November 1998, 81 (4), pp. 608-16. Kiyotaki, Nobuhiro and Wright, Randall. “On Money as a Medium of Exchange.” Journal of Political Economy , August 1989, 97 (4), pp. 927-54. Lagos, Ricardo and Wright, Randall. “A Unified Framework for Monetary Theory and Monetary Analysis.” Journal of Political Economy , June 2005, 113 (3), pp. 463-84.

26 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Lubick, Thomas A. and Schorfheide, Frank. “Testing for Indeterminacy: An Application to U.S. Monetary Policy.” American Economic Review , March 2004, 94 (1), pp. 190-217. Maisel, J. Sherman. Managing the Dollar . New York: W.W. Norton, 1973.

McConnell, Margaret M. and Pérez-Quirós, Gabriel. “Output Fluctuations in the United States: What Has Changed Since the Early 1980’s?” American Economic Review , December 2000, 90 (5), pp. 1464-76.

Meltzer, Allan H. A History of the Federal Reserve. Volume 2, Books 1 and 2. Chicago: University of Chicago Press, 2010.

Meyer, Laurence H. A Term at the Fed: An Insider’s View . New York: Harper Collins, 2004.

Orphanides, Athanasios. “Monetary Policy Rules and the Great Inflation.” American Economic Review , May 2002, 92 (2), pp. 115-20.

Romer, Christina D. and Romer, David H. “Does Monetary Policy Matter? A New Test in the Spirit of Friedman and Schwartz,” in Olivier J. Blanchard and Stanley Fischer, eds., NBER Macroeconomics Annual 1989 . Volume 4. Cambridge, MA: MIT Press, 1989, pp. 121-70.

Romer, Christina D. and Romer, David H. “A Rehabilitation of Monetary Policy in the 1950’s.” American Economic Review , May 2002a, 92 (2), pp. 121-27.

Romer, Christina D. and Romer, David H. “The Evolution of Economic Understanding and Postwar Stabilization Policy.” Presented at a symposium sponsored by the Federal Reserve Bank of Kansas City, “Rethinking Stabilization Policy,” Jackson Hole, Wyoming, August 29-31, 2002b, pp. 11-78; www.kc.frb.org/PUBLICAT/SYMPOS/2002/pdf/S02RomerandRomer.pdf.

Romer, Christina D. and Romer, David. “Choosing the Federal Reserve Chair: Lessons from History.” Journal of Economic Perspectives , Winter 2004, 18 (1), pp. 129-62.

Samuelson, Paul A. and Solow, Robert M. “Analytical Aspects of Anti-Inflation Policy.” American Economic Review , May 1960, 50 (2), pp. 177-94.

Sargent, Thomas J. “Evolution and Intelligent Design.” American Economic Review , March 2008, 98 (1), pp. 5- 37.

Schmitt-Grohé, Stephanie and Uribe, Martín. “Optimal Fiscal and Monetary Policy in a Medium-Scale Macroeconomic Model,” in Mark Gertler and Kenneth Rogoff, eds., NBER Macroeconomics Annual 2005 , Cambridge, MA: MIT Press, 2006, pp. 382-425; www.nber.org/books/gert06-1.

Schreft, Stacey L. “Credit Controls: 1980.” Federal Reserve Bank of Richmond, Economic Review , November/ December 1990, 6, pp. 25-55; www.richmondfed.org/publications/research/economic_review/1990/pdf/er760603.pdf.

Sims, Christopher A. “Price Level Determination in General Equilibrium.” Plenary talk at the Society for Economic Dynamics Annual Meeting, July 2-4, 2009, Istanbul, Turkey.

Sims, Christopher A. and Zha, Tao. “Were There Regime Switches in U.S. Monetary Policy?” American Economic Review , March 2006, 96 (1), pp. 54-81.

Smets, Frank R. and Wouters, Raf. “An Estimated Dynamic Stochastic General Equilibrium Model of the Euro Area.” Journal of the European Economic Association , September 2003, 1(5), pp. 1123-75.

Stock, James H. and Watson, Mark W. “Has the Business Cycle Changed, and Why?” in Mark Gertler and Kenneth Rogoff, eds., NBER Macroeconomics Annual 2002 . Volume 17. Cambridge, MA: MIT Press, 2003, pp. 159- 218.

Stock, James H. and Watson, Mark W. “Why Has U.S. Inflation Become Harder to Forecast?” Journal of Money, Credit, and Banking , February 2007, 39 (S1), pp. 3-33.

FEDERAL RESERVE BANK OF ST . LOUIS REVIEW JULY /AUGUST 2010 27 Fernández-Villaverde, Guerrón-Quintana, Rubio-Ramirez

Tobin, James. “Monetary Policy in 1974 and Beyond.” Brookings Papers on Economic Activity , 1974, 1, pp. 219-32. U.S. Senate. Committee on Banking, Housing, and Urban Affairs. “Nomination of Paul A. Volcker.” Washington, DC: U.S. Government Printing Office, 1979. U.S. Senate. Committee on Banking, Housing, and Urban Affairs. “Nomination of Alan Greenspan.” Washington, DC: U.S. Government Printing Office, 1987. Volcker, Paul A. “The Triumph of Central Banking?” The 1990 Per Jacobbson Lecture, Washington, DC, September 23, 1990; www.perjacobsson.org/lectures/1990.pdf. Volcker, Paul A. and Gyohten, Toyoo. Changing Fortunes: The World’s Money and the Threat to American Leadership. New York: Times Books/Random House, 1992. Wallace, Neil. “Whither Monetary Economics?” International Economic Review , November 2001, 42 (4), pp. 847-69. Weintraub, Robert. “Monetary Policy and Karl Brunner.” Journal of Money, Credit, and Banking , February 1977, 9(1 Part 2), pp. 255-58. Woodford, Michael D. Interest and Prices: Foundations of a Theory of Monetary Policy . Princeton, NJ: Princeton University Press, 2003.

28 JULY /AUGUST 2010 FEDERAL RESERVE BANK OF ST . LOUIS REVIEW